X presented the above electoral map of France with the following comments:
I [X] wonder if [the news article] does not see too much in this picture. First some districts have to be either above or below the national average. Second, the map does not incorporate the population density: very sparsely populated districts in the South-East, like Auvergne or central Corsica are more visible than the densest areas like the Greater Paris, while being more prone to low vaccination rates due to the larger distance to vaccination centres. Third, most of the districts are within ±15% of the average, which may be too large for statistical variation but not much!
I’ve long felt that people who make choropleth maps of France are too unconcerned about artifacts due to population density. I wonder if part of the problem is the apparent uniformity of “the hexagon.” In the U.S. everyone knows there are vast underpopulated areas in the middle of the country, but in France there’s not an equivalent of “the coasts” so it can be easy for people to default to the (wrong) assumption of uniform population density. I wonder if this problem is also made worse by the approximate uniformity in the areas of the départements?
X adds:
Not only the departments are of equivalent size, but the finer administrative division (cantons) is roughly constant in superficy as well, except for the major agglomerations.
This would be an interesting research problem at the intersection of graphical perception and comparative politics, to see how misinterpretations of maps vary across countries.
My recollection is that political geographers struggled a lot with this kind of problem a long time ago, but didn’t come up with any good solutions.
The easiest solution here seems like it would be to add a third dimension — either spatial, tilting the map on its side and have towers stick up and out so volume is proportional to population (e.g. https://i.imgur.com/UShXMSq.png), or maybe adding some other dimension to the colorspace, like opacity (though there already exist light and dark variations on those colors, which would probably make things too busy?).
Instead, wouldn’t the most accepted approach be just to warp geography according to population, a la cartograms? https://en.wikipedia.org/wiki/Cartogram (and I guess shrink estimates of the noisy, low-sample districts w/ a logit-multivariate-normal GP?)
Nikolai:
I’m not a fan of cartograms (this has come up on the blog before, many years ago) because they draw attention to the goofy shapes rather than to the thing being plotted. That said, there are some cartograms without the goofy shapes (I’m thinking of those hex plots of the U.S. where each congressional district is equally sized) which work ok. I think the key is to avoid goofy shapes such as swollen New Jersey etc.
Cartograms are controversial. They were very popular for a while but iirc some variations are hard for laypeople to interpret.
Your multi-dimensional suggestion is a good one, “bivariate choropleth” is the technical name for using two dimensions of color. The other classic solution to this is dasymetric mapping, most commonly using masking based on buildings or other correlates of density to make the density legible. This is much easier to do today than in the past due to data availability and modern GIS.
When Trump first took office, he delighted in showing visitors a map of the U.S. which indicated the 2016 voting outcome in each county in the country. The map was overwhelmingly red with occasional blue even though he did not win the popular vote. An interesting map of that result can be found at
https://www.theatlantic.com/politics/archive/2019/10/trump-2016-election-map-county-population/599335/
According to this article, Trump “won more than 2,600 counties, while Hillary Clinton carried fewer than 500.”
This eye-catching version of the 2016 election has an extra dimension for population density. And other buttons such that you can rotate, enlarge, etc. It is also annoyingly overdone just because the creators could add unnecessary distractions.
Perhaps what you meant to say is that everyone knows that there are vastly overpopulated areas on two of the corners of the continental USA.
Although I tihnk not everyone knows this.
I would say that In France everyone knows that Ile-de-France is much more densely populated that the rest of the hexagon. Ten times as much, in fact. Almost one-fifth of the population of metropolitan France is concentrated in 2% of its territory.
Ile-de-France is the highest density region. However, some would say that the typical French is concentrated far from the HDR…
Carlos:
Sure, people know that. And if you press them on it, they also will know that population is not evenly distributed around the rest of the country either. Still, I suspect there’s a default assumption or partial pooling toward an implicit model of equal populations.
Ok. My point is that in France there is something quite similar to the coasts and the flyover country: la region parisienne et la province.